THE ECONOMETRICS OF MACROECONOMIC MODELLING

8.4 Monetary analysis of Norwegian data

8.4.1 Money demand in Norway—revised and extended data

The demand for broad money in Norway has previously been analysed by Eitrheim (1998) using seasonally unadjusted data from 1969(1) to 1993(4). In that study a cointegrating relationship for money was derived jointly with

cointegrating relationships for wages and consumer prices, and the analysis showed that in the long run, real money balances adjust dynamically to absorb shocks in the real GDP level and the relative price of financial assets (the yield spread) and the relative price of goods (the own real interest rate). In the short run, money balances were also affected by shocks in the exchange rate and private wealth. Evidence for prices being weakly exogenous was also found with respect to the parameters in the money demand relationship, which by implication support the interpretation that it is money holdings that adjust endogenously to changes in the forcing variables in the long run.

In the empirical models in this section we condition on the long-run cointegrating relationship for money balances found in Eitrheim (1998). Assuming homogeneity of degree one in the price level, this relationship can be formulated as:

mt - pt = вуVt + l3rbt(RBt - RTt) + /3rtMP(RTt - &4Pt),

where yt is (log of) real output (GDP), pt is the consumer price index, hence A4pt is the annual rate of headline inflation, RBt is the yield on assets outside money (government bonds with six years maturity), and RTt is the own interest rate on money (the time deposits rate). The yield spread (RBt — RTt) represents the nominal opportunity cost of holding money relative to other financial assets, while the ‘own real interest rate’ (RTt — A4pt) can be interpreted as a measure of the return on money relative to consumer goods.

This long-run equation is grafted into a simplified equilibrium correction model for quarterly money growth with only one lag, which means that (8.5) can be written

Amt = £iAm— + YoAzt + y(Az— + ат(т— - в z—i) + £u et ~ i. i.d.(0,a2). (8.8)

Note that since A4mt = Amt + A3mt-1 we arrive at a relationship for annual money growth A4mt by adding A3mt-1 to both sides of (8.8). If the coefficient on A3mt-1 is close to one, the annual representation is a simple isomorphic transformation of a similar quarterly model.

Re-estimating a money demand model for Norway Compared to Eitrheim (1998), we report results for seven years of new observations. Also, since then, Norwegian National Accounts data for the entire sample period have been substantially revised in order to comply with new international standards, and there has been a major revision in the Monetary Statistics data for broad money holdings.[68] One of the changes in the new definition of broad money is

that unused overdraft facilities and building loans are now excluded. Figure 8.5 shows the revised data along with the data which were analysed in Eitrheim (1998). Despite the exclusion of unused overdraft facilities and building loans, it does not seem that the pattern of annual growth rates in the monetary aggregate has been significantly altered.

Table 8.3 shows the results from re-estimating the model specification in Eitrheim (1998). Despite the revisions of the data for the money and output variables, the old relationship seems to hold up reasonably well on the extended data set. In the short run, money growth is influenced by shocks to the exchange rate (et is (log of) the nominal exchange rate of Norwegian Kroner) and by changes in nominal household wealth (wht). We have also included a dummy variable for the release of tax-exempted savings deposits, M2D914 = 1 in 1991(4) else 0, as well as a variable, S4t * A4RTt, which is intended to pick up the effect from changes in accrued interest earnings, which are capitalised at the end of each year.

Some of the coefficients lose their previous significance, but the re-estimated model passes all mis-specification tests reported in Table 8.3. The estimated <r is 1.13% compared with 0.93% in Eitrheim (1998), so the data fit has deteriorated. From recursive plots (not reported here) of the parameter estimates of the short-run effect from shocks in exchange rates (AAet), it is possible to trace instabilities which may be linked to changes occurring in the exchange rate system in Norway after 1997. After leaving a fixed exchange rate system in 1992 in favour of a managed float, the Norwegian Krone has seen several episodes with more or less free float following speculative attacks, notably in 1997 and 1998. It is not surprising if the currency substitution effect on money holdings did change on those occasions.

Table 8.3

Re-estimating the money demand model for Norway in Eitrheim (1998) on
revised and extended data (seven years of new observations)

Д4 mt = —0.0449(ДДе(_ і + ДДе4_ з) + 0.1383Ди>^_ 2 + 1.0825Дзш^_ і (0.0446) (0.0393) (0.1296)

+ 0.0257 (Дш^ і + Дт^ з) — 0.3107 (Дш^ 2 — Дт^ 4)

(0.1701) (0.0769)

— 0.1026(mt_i — pt-i — 0.8yt-i + 2.25(RB — RM)t-i (0.0197)

— (RM — Д4рЬ_ 1) + 0.0278M2D914t + 0.1505ST * Д4RTt

(0.0120) (0.1623)

+ 0.0186(S1t + S3t) — 0.3756 (0.0037) (0.0718)

<7 = 1.13%

Diagnostic tests

Far(i-S)(5, 114) =1.0610[0.3858]

Farch(i-4)(4, 111) = 1.7918[0.1355]

^normality (2) =0.5735[0.7507]

FHETx2 (16,102) = 1.4379[0.1391]

FReset (1,118) =0.6260[0.4304]

Table 8.4

Improved model for annual money growth, Д4т, for Norway

Д4 mt = —0.0800(ДДе_і + ДДе—з)

(0.0489)

+ 0.1493Sdum97Q1(ДДet_ і + ДДв^ 3)+ 0.1145Дад^_ 2 (0.0886) (0.0367)

+ 1.1134Дзmt_l — 0.3235^mt_2 — Дmt_4)

(0.0394) (0.0464)

—0.1084(mt_i — pt-і — 0.9yt-i + 2.5(RBt_i — RMt-і))

(0.0186)

+ 0.0300M 2D914t + 0.0175(51t + 5 3t) — 0.5272 (0.0111) (0.0021) (0.0898)

<7 = 1.09%

Diagnostic tests

Far(i_5) (5,115) = 0.7026[0.6226]

Farch(i_4)(4, 112) = 0.5574[0.6940]

X2normaiity(2) =2.4736[0.2903]

FHETx2 (14,105) = 1.6997[0.0664]

FReset(1, 119) =0.2022[0.6538]

Note: The sample is 1969(1)-2001(1), quarterly data.

Long run: mt = pt + 0.9yt — 2.5(RB — RM)t + ecmmdt.

model are x2orecast(25) = 36.293[0.0673] and Fchow(25,95) = 1.3452[0.1547].6 Hence, the parameter forecast stability has been improved in the revised money demand model in Table 8.4. In Sections 8.7.3 and 8.7.4, we use the equilibrium correction term, ecmmdt, of Table 8.4 to test for neglected monetary effects in models explaining inflation in Norway.